Case Studies
Computer vision, LLM and multimodal AI systems Khadim Hussain has built, with the stack, the numbers and what was hard.
LLM Fine-tuning
STEM reasoning LLMs: fine-tuning gpt-oss-20b and Qwen3-14B
QLoRA fine-tunes of gpt-oss-20b and Qwen3-14B for STEM reasoning, published as LoRA adapters, merged models and GGUF files for Ollama and llama.cpp.
· 2 min read
Generative AI
Kiffusion and SpaceVision: space image generation with LoRA
A 1,964-pair space image dataset captioned with JoyCaption, and LoRA adapters for FLUX.2-klein (9B) and Kolors, published on Hugging Face.
· 1 min read
Multi-Modal AI
AI Memory: multimodal semantic search for AI agents
A self-hosted memory service that lets AI agents store and search text, images, audio and video, built on CLIP, Whisper, Qdrant and Apache AGE.
· 1 min read